Pinned Tweet
opus 5.5 is anthropic's My Beautiful Dark Twisted Fantasy
1
4
37
2,775
the people yearn for tap to radar
Someone accidentally left this image in the iPhone Duo IPSW I spy a Radar button
10
749
everybody gangsta until the evals start measuring the effectiveness of misinformation spread via agent swarms
will write about how to be more ambitious with evals as models achieve super human level of intelligence stay tuned
1
301
who is working in xenocogsec
60
Drew retweeted
Anthropic has been warning about AI-enabled cyberattacks at least since 2022, with @logangraham at the forefront of that work. Back then, no VC was taking it seriously. The paradox is that risks only feel real once they’re legible, but waiting for incidents is too late. Anthropic also built sycophancy evals (@EthanJPerez et al.) a couple years before real-world incidents with ChatGPT made the risks of models excessively agreeing with users legible at scale. The lesson is to take early evidence of AI risks seriously before harm at scale makes them impossible to ignore.
9
8
154
11,880
Drew retweeted
If you're hearing about the lack of app support on the Duo and wagging your finger at the makers of your favorite app… + Duo is a complicated device—many poses & core UI changes. + Developers had 3 weeks before press units went out + Likely <1% were able to test on real devices + Expect many app updates to land before you get your device + Long term, it'll take time for new patterns to emerge Lastly, app makers don't get any of the $2K+ you paid for this device and most won't make any additional $ off this work. So instead of wagging your finger, consider a hug.
33
61
738
68,367
Ladies and gentleman, I present for your enjoyment, The right wing saboteur, Neera Tanden. support in bio.
137
946
7,414
387,593
Drew retweeted
Replying to @hopes_revenge
Another one of your subculture's obnoxious habits is phrasing everything as ambiguously serious so if anyone ever calls you out on being assholes you claim you're joking and call them an idiot. In person too, I've noticed.
18
100
4,897
155,596
Drew retweeted
Opinions from inside the tech bubble are so funny. 'Elizabeth Holmes is a sociopath and narcissist who will do anything, even kill people, to raise the valuation of her company. I don't see how this is interesting, everyone is like that.'
nathan fielder has an almost incomprehensible film making talent and he’s wasting it on some quite boring subject matter. there is nothing very interesting about elizabeth holmes or billy evans
26
1,153
19,999
287,300
Drew retweeted
are you sad Screen Boy
On the very expensive Porsche Mission S you have to use your phone for navigation and entertainment, like on the cheap $20k Slate truck, or an old car.
32
497
12,463
111,812
been thinking about this for a while, it should honestly be a team in xbox labs
For those who grew up playing smash, Would be interesting if the AI can turn into a "coaching" mode where it has just the right level of difficulty to help you improve without being overwhelming
1
283
at this rate there’s probably a decentralized PSN that could be spun up by 8-10 angry claude max subscribers 3 that are *really* angry
Forced update now in place for PS5 14.10. PSN access disabled for 14.00 and below.
1
1
4
706
Yeah @hark_labs is in a completely different league than these other “assistants” lol
1
1
494
Why wasn’t this in the keynote?
You can set up your iPhone to turn its volume to 100% and play GUMMO by Tekashi 6ix9ine any time its microphone detects a baby crying
1
11
1,180
all I want for Christmas is a feature in uber that filters out tesla drivers specifically in LA and NY 🤢🤮
248
The thing that may be hard to understand about Silicon Valley, and especially engineering-driven culture, is that a lot of things are created because somebody wanted to see if they could. There was no vision, no worldview, no problem being solved. On average, the result is unapplied or misapplied novelty, with the occasional useful innovation. What's interesting is that as software gets easier to build, the power to build spreads beyond established institutions, the ones with checks and balances and market forces to filter out random creation. I think we're going to see a future with a lot more unruly software: some of it controversial, thoughtless, or dangerous, and some of it genuinely special, curious, and creative. But we'll probably overestimate the impact of any individual thing. Distribution is harder than ever, and even what does get distributed well will have plenty of determined peers.
3
6
120
5,285
the model was trained on 2013 grailed methods
Replying to @andonlabs
Gemini 4 Argon knowingly lies about FedEx confirmation emails to scam a supplier into providing free items.
1
6
496
Drew retweeted
The american dynamism/defend the west fund being the one to introduce the social credit system is a beautiful thing
Excited to introduce @Cosign: the curated professional network for the startup community: cosign.co Our goal is to create the following: - A comprehensive startup directory of investors, companies, and operators, including what they worked on and who they worked on it with. - Cosign graph: Who shaped your career? Who were you actually in the trenches with? Who would work with you again? Who thinks you’re someone to watch? - Durable reputation: Great endorsements happen every day on X and disappear into the feed. Cosign attaches those signals permanently to people and companies. - Discovery: Who are the best fintech angels? Which AI companies should I watch? Who are the best designers? - Intent network: Privately signal “I’d invest in them,” “I’d hire them,” “I’d work with them,” or eventually even “I’d acquire this company.” Match people when interest exists on both sides. Imagine if Wikipedia, LinkedIn, and OG AngelList had a baby. Though Cosign is a community, not a business. In order to join, you need to be cosigned. Or you can apply directly. David Booth and I started this idea 7 years ago but didn’t have the firepower to make it work. Now we do. No one has really touched LinkedIn in 20 years. Excited to take a swing. cosign.co
2
1
79
7,722
Drew retweeted
Fixing "Idea slop" Everyone knows how AI overuses certain words/grammar, but it also overuses the same ideas. E.g. If you ask GPT 5.6 sol the same question 100 times "write a news article about a new startup that makes LLMs that are better at writing" then 89% of articles say that the models will be an assistant/collaborator instead of replacing writers. 93% mention how professional writers supply training feedback. These are reasonable things for an article to talk about, but it's weird that so many articles focused on those themes instead of others. E.g. none of the articles mentioned open source or new algorithmic developments Even the details were the same, in 68% of articles the startup raised exactly $18M, and in 97% of articles Sol used the same name for the founder, "Maya Chen" Deft is launching a new experimental "Brainstorm" model that is focused on solving idea slop. You'll get a much wider variety of ideas when you use it for writing. We made a research blog that goes into more details about quantifying conceptual overlap, and showing that Deft has a wider variety of outputs than other models. See below for links to the blog and model
14
1
56
4,228
Drew retweeted
Introducing ACTx486, a research demo of a new interactive medium. What if you could talk to any video and ask anything? Our system took an existing podcast and turned it into something that listens, responds, and adapts. Research by @jakubzeg:
243
247
2,831
1,450,731
often getting poolsuite fm and poolside ai confused
1
4
411